Snowfall Data and Synoptic Fields 1 Snowfall Data and Event Definition

2.3.1a Daily snowfall data

This study parsed daily snowfall data to estimate storm snowfall totals associated with specific synoptic disturbances. Daily snowfall records for the period from 1950 to 2000 served as the source of the snowfall data. These data came from the National Climatic Data Center’s Cooperative Summary of the Day CD-Rom (NCDC 2002), which consists of daily observations of minimum and maximum temperature, precipitation, snowfall, and snow

depth. This study used 121 cooperative observer stations in the Southern Appalachians, which represented all of the stations in the region with a minimum of 10 years of snowfall data.

Since most coop observers are volunteers, the quality of the data can sometimes be a concern, particularly with snowfall and snow depth measurement (Robinson 1989).

However, even when observers follow standardized procedures for measuring snowfall, variation may still exist within the same area due to inter-observer variability (Doesken and Judson 1997, Doesken and Leffler 2000). The observation times of the coop data also introduce some inconsistency for snowfall measurement, as some stations report at 7 AM, others at 5 PM, and a handful at 12 AM. In situations where snowfall events occurred less than 24 hours apart, a small portion of a coop station’s event total snowfall may be attributed to the prior or subsequent event due to different reporting periods. It would be helpful to have fully standardized and higher quality coop data, but unfortunately these are the only snowfall data available at sufficient spatial resolution to fulfill the aims of this research project.

2.3.1b. Event definition using first-order stations

Some periods of snowfall may be connected with a single synoptic-scale disturbance and therefore span only one day, whereas others may result from multiple disturbances or a prolonged period of upslope flow and thus have durations ranging from two to as many as five days. The daily snowfall data by themselves are informative, but are not tied explicitly to synoptic-scale snowfall events. Therefore, it was necessary to calculate event snowfall totals by summing the daily snowfall data through the duration of each event. A total of 1,641

snowfall were identified across the region for the 50-year period, or an average of 32 events per snow season.

Figure 2.3. Topography of the Southern Appalachians with weather stations used in the study.

Each snowfall event was identified by the occurrence of measurable (≥ 0.1 in) snowfall at one or more coop stations across the region. If snowfall was reported, hourly weather conditions (i.e. precipitation occurrence and type), precipitation intensity, and temperature data were acquired for all available first-order stations in the region (Fig. 2.3): Knoxville, TN, Tri-Cities, TN, Beckley, WV, Asheville, NC, and Roanoke, VA. The beginning time of the snowfall event was defined as the hour snow was first reported at any

of the first-order stations, the maturation time the hour in which the spatial extent of snow was the greatest (Fig. 2.4 – i.e. reported at the most number of first-order stations in the same hour), and the ending time the hour snow no longer was reported at any of the first order stations.

Maturation Hour =

Greatest Spatial

Extent of Snow

Two-day

event total

First-order

stations

Coop stations

Figure 2.4. Identification of event maturation hour and calculation of event snowfall totals. Shown are hourly patterns of precipitation type (S = snow, R = rain) and temperature for first-order stations (top) and daily snowfall totals and event snowfall totals for selected coop stations (bottom).

An event remained active as long as precipitation was reported at any first-order station other than Beckley, WV. In Beckley, snowfall often continued for several days due to orographic effects, when in fact different synoptic-scale disturbances in close succession were responsible for the prolonged snowfall. Since the other first-order stations are lower in

elevation and considerably less exposed to northwest flow than Beckley, the orographic effects are much less pronounced. Therefore, event durations are shorter and snowfall can be more easily tied to the individual synoptic-scale disturbances. Therefore, it is possible to break up these prolonged periods of upslope snows in Beckley using data from Tri-Cities and Knoxville. The event total snowfall is thus calculated by summing the daily coop station snowfall data for each day the snowfall event remains active (Fig. 2.4).

A minor deficiency of the first-order time series data is the absence of a high elevation station. Beckley, at 763 m (2,504 ft) and Asheville, at 683 m (2,117 ft) are the highest stations; however, roughly 21 percent of the coop stations exceed these elevations and 11 percent exceed 1000 m (3,280 ft). In some events, NWFS will continue for a longer period in the higher peaks; therefore, the calculated event ending times may be somewhat earlier (and hence event durations somewhat shorter) than is actually the case since NWFS may continue for a longer period along higher elevation windward slopes. In other cases, snowfall may occur at the High Peaks while Asheville and Beckley receive only rain. In fact, approximately 20 percent of the snowfall events are not connected with any hourly snowfall reports at the first order stations. In these cases, the beginning, maturation, and ending times were estimated based on the hourly observations of temperature and precipitation at the first order stations. In a few instances, no hourly precipitation was reported at any of the first- order stations, in which case the beginning, maturation, and ending times were estimated using the daily precipitation data and hourly cloud cover and temperature data.

2.3.1c. Development of snow regions

Due to the large areal extent and significant topographic diversity within the Southern Appalachians, the study area was divided into 14 snow regions. The primary purpose of these groupings was to facilitate intra-regional comparisons. Coop stations were grouped together into snow regions based on similarities in snowfall patterns, elevation, and topography (Table 2.1 and Fig. 2.5). In most cases these regions correspond to zone groupings that the National Weather Service (NWS) uses for forecast products. However, it is important to note that the High Peaks Snow Region (Region 14) is not contiguous, but rather consists of all locations in the study area with elevations greater than 1,200 m (4,000 ft). These areas are largely limited to the mountains of eastern Tennessee and western North Carolina. Recent work (Perry and Konrad 2006) has shown that higher elevation stations (e.g. Mt. Mitchell, NC) in close proximity to valley stations (e.g. Asheville, NC) average as much as 75 cm (1,700 percent) more NWFS on an annual basis, suggesting dramatic differences in snowfall climatologies. Mean and maximum event snowfall totals for those stations reporting snow were calculated for each snow region. The use of snow regions also helps to minimize the deleterious effects of missing or inaccurate data at individual coop stations by placing greater emphasis on regional rather than the local or point scale patterns. Though useful for the purposes of this study, it is important to recognize that considerable variability of snowfall still occurs within the snow regions, particularly in those with greater topographic relief.

Table 2.1. Snow regions and cooperative observer stations.

Snow Region Coop Stations

1. Southern Tennessee Valley Athens, Cleveland, Dayton, Kingston, Knoxville, Lenoir City, Pikeville, Rockwood

2. Southwest Mountains Blairsville, Andrews, Murphy, Copperhill

3. Southern Foothills Clayton, Helen, Toccoa, Forest City, Tryon, Long Creek,

Pickens, Salem

4. Great Smoky Mountains Cataloochee, Hot Springs, Marshall, Oconoluftee, Waterville, Waynesville, Gatlinburg

5. Southern Blue Ridge Asheville, Black Mountain, Brevard, Coweeta, Cullowhee,

Fletcher, Franklin, Highlands, Lake Toxaway, Pisgah Forest, Rosman, Caesars Head

6. Southern Plateau Allardt, Crossville, Newcomb, Norris, Oneida, Tazewell

7. Northern Tennessee Valley Bristol, Elizabethton, Erwin, Greeneville, Kingsport, Rogersville, Abingdon

8. High Country Banner Elk, Blowing Rock, Boone, Celo, Jefferson, Transou,

Trout Dale, Mountain City

9. Central Foothills Lenoir, Marion, Morganton, North Wilkesboro, Yadkinville

10. New River Valley Sparta, Blacksburg, Bland, Floyd, Galax, Meadows of Dan,

Pulaski, Wytheville

11. Northern Foothills Martinsville, Philpott Dam, Roanoke, Rocky Mount, Stuart

12. Central Plateau Baxter, Blackmont, Burdine, Closplint, Cumberland Gap, Pine

Mountain, Skyline, Big Stone Gap, Grundy, John Flanagan, North Fork Reservoir, Pennington Gap, Wise

13. Northern Plateau Burkes Garden, Beckley, Bluefield, Bluestone Lake, Flat Top,

Lewisburg, Pineville, Princeton, Union, White Sulphur Springs

Figure 2.5. Snow regions used in the study. The High Peaks (Region 14) are shaded white on the large map and black on the inset map.

2.3.1d. Identification of NWFS events

NWFS events were defined by the occurrence of a northwest (270 to 360 degrees) low-level (850 hPa) wind at event maturation (hour at which the spatial extent of snow was greatest) (Fig. 2.6). Values for 850-hPa wind direction at the mean center of each snowfall event were calculated from NCEP reanalysis data (discussed in next section) using Synoptic Climatology Suite (Konrad and Meaux 2002) and written to a database file. NWFS events were then selected using the “filter” option in Microsoft Excel. Wind direction at the 850-hPa level (found at approximately 1,450 m (4,757 ft) in periods of northwest flow) is a good indication of the mean wind direction between 1,000 and 2,000 m (3,208 and 6,562 ft), where

much of the orographic enhancement likely occurs. Additionally, the 850-hPa level is a standard height at which a variety of atmospheric data are routinely collected and archived by several governmental agencies and represents a good source of long-term wind direction data. A total of 859 NWFS events were identified across the Southern Appalachians during the period 1950 to 2000, or an average of 17 per year.

Figure 2.6. Identification of NWFS events.

The identification of NWFS events using the automated approach and NCEP reanalysis dataset in the manner described above has many advantages. First of all, it provides a consistent and objective method of identifying NWFS events based on their defining characteristic: low-level (850 hPa) northwest flow. Secondly, values for the 850 hPa

wind direction are extracted at the time in which the spatial extent of snowfall (event maturation) is greatest across the region. Lastly, this automated approach is useful for exceptionally long time series of data, such as the one utilized in this study, where a manual synoptic typing using weather map analysis is unwieldy. One caveat of this methodology is that some of the major snowstorms that affect the Southern Appalachians may be classified as NWFS, even though they are associated with Gulf Lows of the Miller Type A variety (Gurka et al. 1995) with a prolonged period of NWFS on the back side of the surface cyclone. For example, using this methodology, the Blizzard of ’93 exhibited an 850-hPa northwest flow at event maturation, but a substantial amount of the storm total snowfall was also tied to low-level south and southeast flow prior to backing of winds to the northwest. Therefore, the amount of snowfall occurring in conjunction with low-level northwest flow may be over reported. Due to the limited temporal resolution (i.e. 24-hr totals only) of the coop data, it was unfortunately impossible to assign snowfall accumulations to specific low- level wind directions in these events. Nonetheless, even in the extreme snowstorms such as the Blizzard of ’93, the bulk of the total snowfall at higher elevations and windward slopes occurred in conjunction with low-level northwest flow (e.g. Goodge and Hammer 1993).

2.3.2. Synoptic Fields

Table 2.2 lists 16 synoptic fields that were extracted from the National Center for Environmental Prediction (NCEP) reanalysis dataset (Kalnay et al. 1996) and used to characterize the synoptic environment associated with the study events. These 2.5 x 2.5 degree gridded data were generated from a data assimilation routine that incorporated

and therefore provide unmatched continuity over a long period of time. The gridded data were used to obtain the u and v components of 850 hPa wind direction. These data were then spatially interpolated onto a 1776 km grid (197 km resolution), with the center grid

corresponding to the mean center (weighted by coop snowfall totals) of each snowfall event. Using the 0000 and 1200 UTC gridded synoptic fields, a temporal interpolation was

undertaken to estimate field values during the event maturation time. This interpolation procedure was carried out for the remaining synoptic fields. This study used an inverse distance technique to carry out all spatial and temporal interpolations. These fields together provide a good characterization of the synoptic environment (e.g. moisture, circulation, stability, thermal structure, etc.) associated with NWFS.

Table 2.2. Synoptic fields derived from the NCEP reanalysis dataset used in this study. Synoptic Variables Units

1000 hPa Height Meters

500 hPa Height Meters

850 hPa Mixing Ratio g H20/kg dry air

850 hPa Relative Humidity Percent

850 hPa Wind Direction Degrees

850 hPa Wind Speed ms-1

850 hPa Temperature ºC

850 hPa Thermal Advection ºC/12 hour

1000-500 hPa Mean Relative Humidity Percent

700 hPa Vertical Velocity hPa/hour

850-500 hPa Lapse Rate ºC

Precipitable Water Inches

500 hPa Relative Humidity Percent

850 hPa Theta-E ºC

500 hPa Vorticity 1*10-5_s-1

Although the NCEP reanalysis data are unmatched at the synoptic-scale and provide the only available long-term source of data, several minor limitations are worth discussing. The limited spatial resolution (2.5 degrees by 2.5 degrees) is one concern, as mesoscale vorticity maxima and other sub-synoptic-scale features are not readily discernible. Data are also limited to five mandatory levels: 1000, 850, 700, 500, and 200 hPa, when greater vertical resolution, particularly in the lower troposphere, would be helpful. The moist layer, which contributes to NWFS in the Southern Appalachians, is typically confined to the lower troposphere, with the top of the moist layer rarely extending above 600 hPa. Therefore, greater vertical resolution would aid in the assessment of synoptic patterns of humidity, moisture, and temperature within the critical layer of snowfall development. This is also true with some of the other synoptic fields, such as lapse rate, which is only calculated between 850 and 500 hPa. This measure therefore does not adequately capture the finer scale conditional instability in the synoptic environment below 800 hPa characteristic of most NWFS events. Greater temporal resolution of the NCEP reanalysis data would also be desirable, as the temporal interpolation scheme using the inverse distance technique is likely to introduce some error for events in which an abrupt change in the synoptic pattern occurs in conjunction with event maturation times four to six hours before of after the 0/12 Z data. Vertical temperature and moisture profiles will be calculated from atmospheric sounding data to provide a more detailed perspective (see Section 4f below).

2.4. Analyses

2.4.1. Climatology of NWFS

A general climatology of NWFS in the Southern Appalachians was constructed by snow region using the snowfall data from 1950 to 2000. Mean annual snowfall was

calculated for each snow region by summing the mean snowfall of those stations reporting snowfall for each event and then dividing by 50 years. Likewise, average annual NWFS by snow region was calculated in a similar fashion, allowing the percent of average annual snowfall attributed to NWFS to be readily calculated. A negative bias in these calculations may exist, however, as coop stations are largely positioned in valleys as opposed to ridge tops. The NWFS events were stratified into two groups according to the polarity of the synoptic-scale vertical motions at 700 hPa (e.g. events associated with synoptic-scale ascent vs. subsidence). Events were stratified into these two groups in order to differentiate between NWFS events that are primarily due to synoptic-scale disturbances versus those in which orographic forcing predominates. It is well known that a large percentage of NWFS events occur in the absence of synoptic-scale ascent across the Southern Appalachians, but it is unclear how differences in the topographic characteristics of snow regions may influence the frequency or accumulations of NWFS events connected with the two primary types of

forcing. Therefore, the percent of NWFS events in each region connected to a) synoptic-scale ascent and b) synoptic-scale subsidence were calculated. Mean annual snowfall was also calculated for each group and compared across snow regions.

The mean number of NWFS events by year and month for each snow region was also calculated and the results mapped. A NWFS event, by definition, occurred if at least one coop station in a snow region reported measurable snowfall (> 0.1 in, or 0.25 cm). Lastly,

the mean snowfall accumulation associated with each NWFS event was calculated by totaling the snowfall for the 50-year period and then dividing by the number of events. This was done for the entire snow season as well as each month. The purpose of this analysis was to ascertain if any significant differences in mean event snowfall totals existed across the study area.

2.4.2. Intensity of Snowfall

Comparisons were also made between light and heavy NWFS events. Light events may only result in minimal to moderate impacts, whereas the impacts may be considerably greater with heavy events. Therefore, it is helpful to compare the synoptic fields and map patterns between light and heavy events to better understand how and why heavier snowfall occurs in some events, but not in others. Many previous precipitation studies have focused only on heavy events (e.g. Maddox et al. 1979, Mote et al. 1997), when a comparison of the exceptional (heavy) events with the ordinary (light) events is perhaps more useful from an analytical and forecasting standpoint. The snow regions chosen for further analysis were the higher elevation windward slopes: Great Smoky Mountains (Region 4), High Country (Region 8), Central Plateau (Region 12), Northern Plateau (Region 13), and High Elevations (Region 14) (Fig. 2.7). In the other snow regions, the number of NWFS events is

significantly lower, as are the typical snowfall accumulations and associated impacts. The Fields Developer and Climatology Explorer modules of the Synoptic

Climatological Suite (Konrad and Meaux 2002) allowed for NWFS events to be stratified by intensity. Light events were defined as the bottom three quartiles of the mean snowfall of those cooperative observer stations reporting snowfall in the respective snow regions, and

heavy events were defined as the top quartile. The heavy events represent the tail of the histogram (Fig. 2.8). In the High Country (region 8), all of these events met the current criteria for an advisory and over half for a warning, with an areal mean snowfall of at least 5.6 cm (2.2 in) (NWS 2006). Composite mean values for each of the 15 synoptic fields centered over each event were compared between the two groups and the statistical significance evaluated using a two-sample difference of means t-test. Composite plots of selected synoptic fields were also developed for both the light and heavy events to better illustrate the synoptic patterns. In addition, composite map patterns of the mean difference between heavy and light events were developed for those synoptic fields that displayed a significant (p < 0.05) difference.

0 20 40 60 80 100 120 140 1 2 3 4 5 6 7 8 _{9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25} 26+ Snowfall (cm) Numbe r of Events 75th Percentile

Figure 2.8. Frequency of NWFS events by snowfall amount for the High Country.

In addition to light versus heavy comparisons for the entire sample of NWFS events in each of the five snow regions, this study also evaluated snowfall intensity for events grouped by synoptic-scale vertical motions. Snowfall accumulations are typically greater and more widespread (e.g. more likely to occur at lower elevations and along windward slope) in the presence of synoptic-scale ascent, whereas forcing is primarily orographic in situations of synoptic-scale subsidence, with lighter accumulations that are limited to the higher elevations and windward slopes. Therefore, light and heavy events were compared for just those events